Colour the black & white
Trained on millions of dated photographs so the palette matches the era — not the cartoonish saturation other tools produce.
How it knows the colour
The model has seen millions of colour photographs from every decade since 1907, tagged by year, place, and content. When you give it a black-and-white photo, it estimates the most likely colours for the things it recognises: skin, sky, grass, brick, a Coca-Cola sign, a fire engine, a midwestern lawn in August. It does this with statistical confidence, not omniscience. The result is plausible, not certified.
What it gets right, what it does not
Skin tone, sky, foliage, wood, water — usually very close. Clothing colour — sometimes a guess: a navy dress and a black dress look the same in grey. A red car may come back as orange. Take the result as a careful estimate, not a forensic reconstruction.
One honest note
If your photograph documents a person who is no longer with us, you may want to know that what you see is a likeness in their probable colours. We say this because it matters.